Determining linear trend after multiple imputation

[jemima spencer]

Hello,

(Firstly, really sorry for lack of direct export of STATA code - I'm using it via a secure VDI so have had to re-write here)


Could somebody kindly help me with running commands to assess the strength of a linear trend when using multiple imputation with logistic regression? I am looking at odds of coronary angiogram in people with different degrees of kidney impairment (an ordinal variable - ckdstages 1-5).

This was all working for me without multiple imputation. My code was:

logistic angiogram i.ckdstage i.ageyears_grp obese cvd diabetes...

contrast p.ckdstage, noeffects


************************************************** ************************************************** *****************************
However, I then realised I needed to impute some ethnicity data and using multiple imputation for this.

mi estimate, or post: logit angiogram i.ckdstage i.ageyears_grp obese cvd diabetes...


I think I need to now use the command mimrgns to assess trend, but have failed to make head or tails of the STATA file on this command. Could somebody possibly help me?



Sorry if this is terribly obvious. It's sadly not to me!

Thankyou so much for your time.

Jemima

[Tiago Pereira]

I am unclear why the independent variable is used as a categorical variable, and then you want to check the linear trend.

have you tried - mi test- ?

https://www.statalist.org/forums/for...er-mi-estimate

[jemima spencer]

Thankyou.

The independent variable ckdstage is an ordinal variable with five categories, ie worsening kidney impairment. I want to see what the strength of the linear trend is for odds of angiography between the different levels of kidney impairment.

I had a look at mi test, but I'm not sure this assess linear trend does it?

Thanks again

Jemima

[daniel klein]

Not commenting on the substantial question, technically you probably want

mimrgns p.ckdstage , contrast(pveffects) asbalanced

The reported z-tests are equivalent to the Chi2-tests that you obtain from contrast.

mimrgns is most likely from SSC (or GitHub).

[jemima spencer]

Thankyou so much! I'll try this now.

[jemima spencer]

Dear Daniel,

Worked perfectly. Thankyou